function [PRED_tv_var, PB,outR, outy2star,outB,outA,PR] = TvVarEstKSC(data,tight_Q,tight_W,tight_S,gbit,maxgbit,burn,jump,post_index,trs,trsMax,MaxEig,MaxRoot,c,L,horizon,ksc);
% Estimation of a Time-Varying Coefficients Stochastic Volatility VAR using
% Kim Shephard and Chibb (1998) algortithm.
VVV=0.2;
KKK=13;
ANTO=3;
%%%%%%%%%%%%%%%%%%%%%%% Multivariate %%%%%%%%%%%%%%%%%%%%%%%%%%%%
if size(data,2)>1
    Y=data;


    %%%% Set parameters
    T0=KKK-L; % period in the initial sample
    [yy xx] = VAR_str(Y(1:end,:),c,L);     % yy and XX all the sample
    [TT r]=size(yy);
    for t=1:TT
        XX(:,:,t)=kron(eye(r),xx(t,:)');
    end
    y=yy(T0+1:end,:);
    x=xx(T0+1:end,:);
    X=XX(:,:,T0+1:end);
    % NEW LINES FROM THE BEGINNING OF THE SAMPLE
    %     y=yy(1:end,:);
    %     x=xx(1:end,:);
    %     X=XX(:,:,1:end);


    r=size(y,2);
    [T rp] = size(x);
    %%%% OLS in the pre sample
    [B0,VB0,R0] = sur(yy',XX,T0);
    resid=innovm(yy(1:T0,:)',xx(1:T0,:),B0*ones(1,T0),r,T0,L)'; % resuduals

    %%%% Variance and mean of theta0
    VB0=VVV*VB0;

    %%%% Variance and mean of log(sigma0)
    H0=log(diag(sqrt(R0)));
    VH0=sqrt(10)*eye(r);%diag(diag(R0))*2;

    %%%% Mean and variance of alpha0
    cR0=chol(R0)';
    ccR0=cR0*inv(diag(diag(cR0)));
    A0=reshape((tril(ccR0)-diag(diag(ccR0)))',r^2,1);
    A0=A0(find(A0));
    co1=[];
    co2=[];
    for i=1:r-1
        dd=size(co1,1)+1:size(co1,1)+size(co2,1)+1;
        VA0(dd,dd) = inv(resid(:,1:i)'*resid(:,1:i))*diag(R0(i+1,i+1));
        co2=ones(i,1);
        co1=[co1;co2];
    end
    VA0=VA0*4;

    %%%% Scale matrix for the variance in covariances innovations
    U0=tight_S*VA0/4;%0.001*VA0/4;
    %U0=0.001*VA0/4;
    dfU0=[2:r];

    %%%% Scale matrix for the variance in coefficients innovations
    %Q0 = 0.0001*VB0;

    %Q0 = tight_Q*VB0;
    Q0 = tight_Q.*VB0;
    df0 = T0;%size(B0,1)+1;%T0; %
    TQ0=df0*Q0;
    df = T+df0;

    %%%% Scale matrix for the variance in volatilities innovations
    W0=tight_W*eye(r);%0.001*eye(r);
    %W0=0.001*eye(r);
    dfW0=r+1;

    %%%% Initialize Yhat, ystar,y2star
    dy = diff(y(:,1:r));
    e(:,1:r) = dy - ones(T-1,1)*mean(dy);
    Yhat=[zeros(r,1) e']';
    InitS=cov(Yhat);
    InitA=chol(InitS)';
    ystar=inv(InitA*inv(diag(sqrt(diag(InitS)))))*Yhat';
    y2star=log(ystar'.^2+0.001);

    %%%% Initialize states
    for i=1:r
        for t=1:T
            for j=1:7
                probs(j)=ksc(j,2)*normpdf(y2star(t,i),y2star(t,i)+ksc(j,3)-1.2704,ksc(j,4));
            end
            s0(i,t)=DiscreteDraw(probs./sum(probs));
        end
    end
    s=s0;

    %%%% Initialize variances
    U=U0;
    W=W0;
    Q=Q0;
    %R=R0;

    %%%% Coefficitent matrix in the state space for volatilities
    h=eye(r)*2;

    %%%%%%%% Initial draws: this is useful for the last part of the program
    P1(:,:,1) = VH0; % P(1|0)
    K = (P1(:,:,1))*h'*inv(h* P1(:,:,1)*h' + diag(ksc(s(:,1),4))); % K(1)
    P0(:,:,1) = P1(:,:,1) - K*h*P1(:,:,1); % P(1|1)
    S0(:,1) = H0 + K*( y2star(1,:)' - h*H0 - (ksc(s(:,1),3)-1.2074)); % S(1|1)
    for i = 2:T,
        P1(:,:,i) = P0(:,:,i-1) + W; % P(t|t-1)
        K = (P1(:,:,i))*h'*inv(h*P1(:,:,i)*h' + diag(ksc(s(:,i),4))); % K(t)
        P0(:,:,i) = P1(:,:,i) - K*h*(P1(:,:,i)); % P(t|t)
        S0(:,i) = S0(:,i-1) + K*( y2star(i,:)' - h*S0(:,i-1) - (ksc(s(:,i),3)-1.2704) ); % S(t|t)
    end
    wa = randn(r,T);
    SA(:,T) = S0(:,T) + real(sqrtm(P0(:,:,T)))*wa(:,T);
    for i = 1:T-1,
        PM = P0(:,:,T-i)*inv(P1(:,:,T-i+1));
        P = P0(:,:,T-i) - PM*P0(:,:,T-i);
        SM = S0(:,T-i) + PM*(SA(:,T-i+1) - S0(:,T-i));
        SA(:,T-i) = SM + real(sqrtm(P))*wa(:,T-i);
    end
    lh=SA';
    H=exp(2*SA'); %% compute variance fr

    %%%% 2) Draw covariances
    TA=zeros(r*(r-1)/2,r*(r-1)/2);
    A=[];
    AA=[];
    for i=1:r-1
        dd=size(A,1)+1:size(A,1)+size(AA,1)+1;
        [AS0,AP0,AP1] = kfR1RegTvc(Yhat(:,i+1)',-Yhat(:,1:i)',U(dd,dd),H(:,i+1),A0(dd),VA0(dd,dd),T);
        AA = gibbs1Reg(Yhat(:,i+1),-Yhat(:,1:i),AS0,AP0,AP1,T,length(dd));
        [TU(dd,dd),dfU] = iwpQ(AA,r,T,L,dfU0(i)*U0(dd,dd),dfU0(i));
        A=[A;AA];
        g=real(sqrtm(inv(TU(dd,dd))));
        ug=randn(i,dfU);
        U(dd,dd) = inv(g*ug*ug'*g');
        CH=A;
        clear AS0 AP0 AP1;
    end
    for t=1:T
        CF(:,:,t)= chofac(r,A(:,t));
        R(:,:,t) = inv(CF(:,:,t))*diag(H(t,:))*inv(CF(:,:,t))';
    end

    for t=1:T
        CF(:,:,t)= chofac(r,A(:,t));
        R0(:,:,t) = inv(CF(:,:,t))*diag(H(t,:))*inv(CF(:,:,t))';
    end
    R=R0;

    %%%% initial draw of states for AR coefficients
    [SS0,PP0,PP1] = kfR1(y',X,Q0,R,B0,VB0,T,r,L);

    while MaxRoot>MaxEig

        %[S0,P0,P1] = kfR(y',X,Q0,CF,H,B0,VB0,T,r,L);
        th_backward0 = gibbs1(y',X,SS0,PP0,PP1,T,r,L);
        %[th_backward0,P_backward0] = tvc_backward_filter(y,X,B0,VB0,R0,Q0,0);
        for j =1:T
            comp_mat = companion(th_backward0(:,j),L,r,c);
            AutVal(j) =  max(abs(eig(comp_mat)));
        end
        MaxRoot=max(AutVal)
    end
    B=th_backward0;
    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

    %%%% Initialize matrices of draws
    PQ=zeros(rp*r,rp*r,(maxgbit-burn)/jump);
    PR=zeros(r,r,T,(maxgbit-burn)/jump);
    PB=zeros(rp*r,T,(maxgbit-burn)/jump);
    PW=zeros(r,r,(maxgbit-burn)/jump);
    PU=zeros(r*(r-1)/2,r*(r-1)/2,(maxgbit-burn)/jump);
    Ps=zeros(r,T,(maxgbit-burn)/jump);
    PA=zeros(r*(r-1)/2,T,(maxgbit-burn)/jump);

    PQ(:,:,post_index)=Q;
    PR(:,:,:,post_index)=R;
    PB(:,:,post_index)=B;
    PW(:,:,post_index)=W;
    PU(:,:,post_index)=U;
    Ps(:,:,post_index)=s;
    PA(:,:,post_index)=A;


    %%%% Start the chain
    disp('begin MCMC')
    tic,
anto = 0;
    while gbit<=maxgbit

        itnum=num2str(gbit);
        if itnum(size(itnum,2))==num2str(0) %& itnum(size(itnum,2)-1)==num2str(0)
            disp(['Iteration   ' num2str(gbit) '    Collected Draws   ' num2str(post_index)...
                '   Done   ' num2str(gbit/maxgbit*100) '%    MaxEig   ' num2str(max(AutVal)) ...
                '   Elapsed time ',num2str(toc/60),' mnts'])
        end

        %%%% 1) Draw stochastic volatilities
        P1(:,:,1) = VH0; % P(1|0)
        K = (P1(:,:,1))*h'*inv(h* P1(:,:,1)*h' + diag(ksc(s(:,1),4))); % K(1)
        P0(:,:,1) = P1(:,:,1) - K*h*P1(:,:,1); % P(1|1)
        S0(:,1) = H0 + K*( y2star(1,:)' - h*H0 - (ksc(s(:,1),3)-1.2074)); % S(1|1)
        for i = 2:T,
            P1(:,:,i) = P0(:,:,i-1) + W; % P(t|t-1)
            K = (P1(:,:,i))*h'*inv(h*P1(:,:,i)*h' + diag(ksc(s(:,i),4))); % K(t)
            P0(:,:,i) = P1(:,:,i) - K*h*(P1(:,:,i)); % P(t|t)
            S0(:,i) = S0(:,i-1) + K*( y2star(i,:)' - h*S0(:,i-1) - (ksc(s(:,i),3)-1.2704) ); % S(t|t)
        end
        wa = randn(r,T);
        SA(:,T) = S0(:,T) + real(sqrtm(P0(:,:,T)))*wa(:,T);
        for i = 1:T-1,
            PM = P0(:,:,T-i)*inv(P1(:,:,T-i+1));
            P = P0(:,:,T-i) - PM*P0(:,:,T-i);
            SM = S0(:,T-i) + PM*(SA(:,T-i+1) - S0(:,T-i));
            SA(:,T-i) = SM + real(sqrtm(P))*wa(:,T-i);
        end
        lh=SA';
        H=exp(2*SA'); %% compute variance fr

        %%%% 2) Draw covariances
        TA=zeros(r*(r-1)/2,r*(r-1)/2);
        A=[];
        AA=[];
        for i=1:r-1
            dd=size(A,1)+1:size(A,1)+size(AA,1)+1;
            [AS0,AP0,AP1] = kfR1RegTvc(Yhat(:,i+1)',-Yhat(:,1:i)',U(dd,dd),H(:,i+1),A0(dd),VA0(dd,dd),T);
            AA = gibbs1Reg(Yhat(:,i+1),-Yhat(:,1:i),AS0,AP0,AP1,T,length(dd));
            [TU(dd,dd),dfU] = iwpQ(AA,r,T,L,dfU0(i)*U0(dd,dd),dfU0(i));
            A=[A;AA];
            g=real(sqrtm(inv(TU(dd,dd))));
            ug=randn(i,dfU);
            U(dd,dd) = inv(g*ug*ug'*g');
            CH=A;
            clear AS0 AP0 AP1;
        end
        for t=1:T
            CF(:,:,t)= chofac(r,A(:,t));
            R(:,:,t) = inv(CF(:,:,t))*diag(H(t,:))*inv(CF(:,:,t))';
        end


        %%%% Draw VAR coefficients
        [SS0,PP0,PP1] = kfR1(y',X,Q,R,B0,VB0,T,r,L);
        B = gibbs1(y',X,SS0,PP0,PP1,T,r,L);


        %%%% compute new Yhat
        Yhat=innovm(y',x,B,r,T,L)';

        %%%% Compute new ystar
        for t=1:T
            ystar(:,t)=CF(:,:,t)*Yhat(t,:)';
        end

        %%%% Compute new y2star
        y2star=log(ystar'.^2+0.001);

        %%%% Compute new probabilities
        for i=1:r
            for t=1:T
                for j=1:7
                    probs(j)=ksc(j,2)*normpdf(y2star(t,i),2*SA(i,t)+ksc(j,3)-1.2704,ksc(j,4));
                end
                s(i,t)=DiscreteDraw(probs./sum(probs));
            end
        end

        %%%% Draw SV innovation variance
        [TW,dfW] = iwpQ(SA,r,T,L,dfW0*W0,dfW0);
        W = gibbs2Q(TW,dfW,r,0);

        %%%% Draw coefficients innovations variance
        [TQ,df] = iwpQ(B,r,T,L,TQ0,df0);
        Q = gibbs2Q(TQ,df,r,L);

        %%%% Check roots
        rootInd=1;
        timInd=1;

        while rootInd==1 & timInd<=T

            comp_mat = companion(B(:,timInd),L,r,c);
            AutVal =  max(abs(eig(comp_mat)));
            if max(AutVal)>MaxEig & trs<trsMax
                rootInd=0;
                trs=trs+1;
                
            elseif max(AutVal)>MaxEig & trs==trsMax
                Q=PQ(:,:,post_index);
                R=PR(:,:,:,post_index);
                B=PB(:,:,post_index);
                W=PW(:,:,post_index);
                U=PU(:,:,post_index);
                s=Ps(:,:,post_index);
                A=PA(:,:,post_index);
                Yhat=innovm(y',x,B,r,T,L)';
                for t=1:T
                    ystar(:,t)=chofac(r,A(:,t))*Yhat(t,:)';
                end
                y2star=log(ystar'.^2+0.001);
                rootInd=0;
                trs=0;

                anto = anto + size(max(AutVal),1);
                if anto>=ANTO;
                     rootInd=1;
                    timInd=timInd+1;
                    %gbit = gbit+1;
                    anto = 0;
                end
                
            elseif max(AutVal)<MaxEig
                timInd=timInd+1;
                anto = 0;
            end
        end

        % sample from markov chain
        if timInd==T+1
            %B=th_b;
            if gbit <= burn
                post_index=1;
                PQ(:,:,post_index)=Q;
                PR(:,:,:,post_index)=R;
                PB(:,:,post_index)=B;
                PW(:,:,post_index)=W;
                PU(:,:,post_index)=U;
                Ps(:,:,post_index)=s;
                PA(:,:,post_index)=A;
            elseif gbit > burn
                check_index = (gbit-burn)/jump- fix((gbit-burn)/jump);
                if check_index == 0
                    post_index = (gbit-burn)/jump+1;
                end
                PQ(:,:,post_index)=Q;
                PR(:,:,:,post_index)=R;
                PB(:,:,post_index)=B;
                PW(:,:,post_index)=W;
                PU(:,:,post_index)=U;
                Ps(:,:,post_index)=s;
                PA(:,:,post_index)=A;
            end
            gbit = gbit+1;
        end
    end

    for i=1:size(PB,3)
        C=[PB(1:L*r+1:end,end,i); zeros(r*(L-1),1)];
        coeff = reshape(squeeze(PB(setdiff(1:r*(r*L+1),1:L*r+1:r*(r*L+1)),end,i)),L*r,r)';
        comp = [coeff; [kron(eye(r),eye(L-1)) zeros(r*(L-1),r)]];
        cons=zeros(size(C,1),1);
        for hh = 1:horizon
            cons = cons+comp^(hh-1)*C;

            pred(:,hh) = cons + comp^hh*reshape(flipud(data(end-L+1:end,:))',1,L*r)';
        end

        PRED_tv_var(:,:,i) = pred(1:r,:);
        
    end

    outR=squeeze(median(PR,4));
    Yhat=innovm(y',x,median(PB,3),r,T,L)';
    for t=1:T
        ystar(:,t)=chofac(r,median(PA(:,t,:),3))*Yhat(t,:)';
    end
    outy2star=log(ystar'.^2+0.001);
    outB=median(B,3);
    outA=median(PA,3);


    %%%%%%%%%%%%%%%%%%%%%%% Univariate %%%%%%%%%%%%%%%%%%%%%%%%%%%%
elseif size(data,2)==1

    PU=[];

    Y=data;

    %%%% Set parameters
    T0=KKK-L; % period in the initial sample
    [yy xx] = VAR_str(Y(1:end,:),c,L);     % yy and XX all the sample
    [TT r]=size(yy);
    for t=1:TT
        XX(:,:,t)=kron(eye(r),xx(t,:)');
    end
    y=yy(T0+1:end,:);
    x=xx(T0+1:end,:);
    X=XX(:,:,T0+1:end);
    % NEW LINES FROM THE BEGINNING OF THE SAMPLE
    %     y=yy(1:end,:);
    %     x=xx(1:end,:);
    %     X=XX(:,:,1:end);
    r=size(y,2);
    [T rp] = size(x);
    %%%% OLS in the pre sample
    [B0,VB0,R0] = sur(yy',XX,T0);
    %resid=innovm(yy(1:T0,:)',xx(1:T0,:),B0*ones(1,T0),r,T0,L)'; % resuduals

    %%%% Variance and mean of theta0
    VB0=VVV*VB0;

    %%%% Variance and mean of log(sigma0)
    H0=log(diag(sqrt(R0)));
    VH0=sqrt(10)*eye(r);%diag(diag(R0))*2;

    %%%% Scale matrix for the variance in coefficients innovations
    %Q0 = 0.0001*VB0;
    %Q0 = tight_Q*VB0;
    Q0 = tight_Q.*VB0;
    df0 = T0;%size(B0,1)+1;%T0; %
    TQ0=df0*Q0;
    df = T+df0;

    %%%% Scale matrix for the variance in volatilities innovations
    W0=0.001*eye(r);
    dfW0=r+1;

    %%%% Initialize Yhat, ystar,y2star
    dy = diff(y(:,1:r));
    e(:,1:r) = dy - ones(T-1,1)*mean(dy);
    Yhat=[zeros(r,1) e']';
    ystar=Yhat';
    y2star=log(ystar'.^2+0.001);

    %%%% Initialize states
    for i=1:r
        for t=1:T
            for j=1:7
                probs(j)=ksc(j,2)*normpdf(y2star(t,i),y2star(t,i)+ksc(j,3)-1.2704,ksc(j,4));
            end
            s0(i,t)=DiscreteDraw(probs./sum(probs));
        end
    end
    s=s0;

    %%%% Initialize variances
    W=W0;
    Q=Q0;
    %R=R0;

    %%%% Coefficitent matrix in the state space for volatilities
    h=eye(r)*2;

    %%%%%%%% Initial draws: this is useful for the last part of the program
    P1(:,:,1) = VH0; % P(1|0)
    K = (P1(:,:,1))*h'*inv(h* P1(:,:,1)*h' + diag(ksc(s(:,1),4))); % K(1)
    P0(:,:,1) = P1(:,:,1) - K*h*P1(:,:,1); % P(1|1)
    S0(:,1) = H0 + K*( y2star(1,:)' - h*H0 - (ksc(s(:,1),3)-1.2074)); % S(1|1)
    for i = 2:T,
        P1(:,:,i) = P0(:,:,i-1) + W; % P(t|t-1)
        K = (P1(:,:,i))*h'*inv(h*P1(:,:,i)*h' + diag(ksc(s(:,i),4))); % K(t)
        P0(:,:,i) = P1(:,:,i) - K*h*(P1(:,:,i)); % P(t|t)
        S0(:,i) = S0(:,i-1) + K*( y2star(i,:)' - h*S0(:,i-1) - (ksc(s(:,i),3)-1.2704) ); % S(t|t)
    end
    wa = randn(r,T);
    SA(:,T) = S0(:,T) + real(sqrtm(P0(:,:,T)))*wa(:,T);
    for i = 1:T-1,
        PM = P0(:,:,T-i)*inv(P1(:,:,T-i+1));
        P = P0(:,:,T-i) - PM*P0(:,:,T-i);
        SM = S0(:,T-i) + PM*(SA(:,T-i+1) - S0(:,T-i));
        SA(:,T-i) = SM + real(sqrtm(P))*wa(:,T-i);
    end
    lh=SA';
    H=exp(2*SA'); %% compute variance fr

    %%%% 2) Draw covariances
    R = H;

    %%%% initial draw of states for AR coefficients

    [SS0,PP0,PP1] = kfR1RegTvc(y',X,Q0,R,B0,VB0,T);
    while MaxRoot>MaxEig
        %[S0,P0,P1] = kfR(y',X,Q0,CF,H,B0,VB0,T,r,L);
        th_backward0 = gibbs1(y',X,SS0,PP0,PP1,T,r,L);
        %[th_backward0,P_backward0] = tvc_backward_filter(y,X,B0,VB0,R0,Q0,0);
        for j =1:T
            comp_mat = companion(th_backward0(:,j),L,r,c);
            AutVal(j) =  max(abs(eig(comp_mat)));
        end
        MaxRoot=max(AutVal)
    end
    B=th_backward0;
    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

    %%%% Initialize matrices of draws
    PQ=zeros(rp*r,rp*r,(maxgbit-burn)/jump);
    PR=zeros(T,(maxgbit-burn)/jump);
    PB=zeros(rp*r,T,(maxgbit-burn)/jump);
    PW=zeros(r,r,(maxgbit-burn)/jump);
    Ps=zeros(r,T,(maxgbit-burn)/jump);

    PQ(:,:,post_index)=Q;
    PR(:,post_index)=R;
    PB(:,:,post_index)=B;
    PW(:,:,post_index)=W;
    Ps(:,:,post_index)=s;


    %%%% Start the chain
    disp('begin MCMC')
    tic,
    anto = 0;
    while gbit<=maxgbit
        itnum=num2str(gbit);
        if itnum(size(itnum,2))==num2str(0) %& itnum(size(itnum,2)-1)==num2str(0)
            disp(['Iteration   ' num2str(gbit) '    Collected Draws   ' num2str(post_index)...
                '   Done   ' num2str(gbit/maxgbit*100) '%    MaxEig   ' num2str(max(AutVal)) ...
                '   Elapsed time ',num2str(toc/60),' mnts'])
        end

        %%%% 1) Draw stochastic volatilities
        P1(:,:,1) = VH0; % P(1|0)
        K = (P1(:,:,1))*h'*inv(h* P1(:,:,1)*h' + diag(ksc(s(:,1),4))); % K(1)
        P0(:,:,1) = P1(:,:,1) - K*h*P1(:,:,1); % P(1|1)
        S0(:,1) = H0 + K*( y2star(1,:)' - h*H0 - (ksc(s(:,1),3)-1.2074)); % S(1|1)
        for i = 2:T,
            P1(:,:,i) = P0(:,:,i-1) + W; % P(t|t-1)
            K = (P1(:,:,i))*h'*inv(h*P1(:,:,i)*h' + diag(ksc(s(:,i),4))); % K(t)
            P0(:,:,i) = P1(:,:,i) - K*h*(P1(:,:,i)); % P(t|t)
            S0(:,i) = S0(:,i-1) + K*( y2star(i,:)' - h*S0(:,i-1) - (ksc(s(:,i),3)-1.2704) ); % S(t|t)
        end
        wa = randn(r,T);
        SA(:,T) = S0(:,T) + real(sqrtm(P0(:,:,T)))*wa(:,T);
        for i = 1:T-1,
            PM = P0(:,:,T-i)*inv(P1(:,:,T-i+1));
            P = P0(:,:,T-i) - PM*P0(:,:,T-i);
            SM = S0(:,T-i) + PM*(SA(:,T-i+1) - S0(:,T-i));
            SA(:,T-i) = SM + real(sqrtm(P))*wa(:,T-i);
        end
        lh=SA';
        H=exp(2*SA'); %% compute variance fr

        %%%% 2) Draw covariances
        R=H;


        %%%% Draw VAR coefficients
        [SS0,PP0,PP1] = kfR1RegTvc(y',X,Q0,R,B0,VB0,T);
        B = gibbs1(y',X,SS0,PP0,PP1,T,r,L);


        %%%% compute new Yhat
        Yhat=innovm(y',x,B,r,T,L)';
        ystar=Yhat';
        %%%% Compute new y2star
        y2star=log(ystar'.^2+0.001);

        %%%% Compute new probabilities
        for i=1:r
            for t=1:T
                for j=1:7
                    probs(j)=ksc(j,2)*normpdf(y2star(t,i),2*SA(i,t)+ksc(j,3)-1.2704,ksc(j,4));
                end
                s(i,t)=DiscreteDraw(probs./sum(probs));
            end
        end

        %%%% Draw SV innovation variance
        [TW,dfW] = iwpQ(SA,r,T,L,dfW0*W0,dfW0);
        W = gibbs2Q(TW,dfW,r,0);

        %%%% Draw coefficients innovations variance
        [TQ,df] = iwpQ(B,r,T,L,TQ0,df0);
        Q = gibbs2Q(TQ,df,r,L);

        %%%% Check roots
        rootInd=1;
        timInd=1;
        while rootInd==1 & timInd<=T
            comp_mat = companion(B(:,timInd),L,r,c);
            AutVal =  max(abs(eig(comp_mat)));
            if max(AutVal)>MaxEig & trs<trsMax
                rootInd=0;
                trs=trs+1;
            elseif max(AutVal)>MaxEig & trs==trsMax
                Q=PQ(:,:,post_index);
                R=PR(:,post_index);
                B=PB(:,:,post_index);
                W=PW(:,:,post_index);
                s=Ps(:,:,post_index);
                Yhat=innovm(y',x,B,r,T,L)';
                ystar=Yhat';
                y2star=log(ystar'.^2+0.001);
                rootInd=0;
                trs=0;

                anto = anto + size(max(AutVal),1);
                if anto>=ANTO;
                     rootInd=1;
                    timInd=timInd+1;
                    %gbit = gbit+1;
                    anto = 0;
                end
            elseif max(AutVal)<MaxEig
                timInd=timInd+1;
                anto = 0;
            end
        end

        % sample from markov chain
        if timInd==T+1
            %B=th_b;
            if gbit <= burn
                post_index=1;
                PQ(:,:,post_index)=Q;
                PR(:,post_index)=R;
                PB(:,:,post_index)=B;
                PW(:,:,post_index)=W;
                Ps(:,:,post_index)=s;
            elseif gbit > burn
                check_index = (gbit-burn)/jump- fix((gbit-burn)/jump);
                if check_index == 0
                    post_index = (gbit-burn)/jump+1;
                end
                PQ(:,:,post_index)=Q;
                PR(:,post_index)=R;
                PB(:,:,post_index)=B;
                PW(:,:,post_index)=W;
                Ps(:,:,post_index)=s;
            end
            gbit = gbit+1;
        end
    end

    for i=1:size(PB,3)
        C=[PB(1:L*r+1:end,end,i); zeros(r*(L-1),1)];
        coeff = reshape(squeeze(PB(setdiff(1:r*(r*L+1),1:L*r+1:r*(r*L+1)),end,i)),L*r,r)';
        comp = [coeff; [kron(eye(r),eye(L-1)) zeros(r*(L-1),r)]];
        cons=zeros(size(C,1),1);
        for hh = 1:horizon
            cons = cons+comp^(hh-1)*C;

            pred(:,hh) = cons + comp^hh*reshape(flipud(data(end-L+1:end,:))',1,L*r)';
        end

        PRED_tv_var(:,:,i) = pred(1:r,:);
    end
    outR=squeeze(median(PR,2));
    Yhat=innovm(y',x,median(PB,3),r,T,L)';
    ystar=Yhat';
    outy2star=log(ystar'.^2+0.001);
    outB=median(B,3);
    outA=[];
end
